CAHLR/pyBKT Title: Consulting opportunity: BKT production integration ($500-1500)

[eraschke66]

Hi,
I'm building PassAI, an EdTech platform that predicts exam pass probability for IB/AP/GCSE students. We have a working Python/FastAPI backend with quiz generation and basic progress tracking, but need proper BKT implementation.

pyBKT is exactly what we've been referencing in our technical spec. We need someone familiar with the library (or the underlying model) to help us integrate production-grade BKT into our system.

The twist: We use Claude API to score quiz responses on a 0-100% scale rather than binary correct/incorrect. So we need someone who can adapt the standard 4-parameter BKT model to handle continuous scoring inputs.

Scope: Review existing code, implement proper BKT with our scoring approach, write tests, 5-day sprint.
Budget: $500-$1,500 fixed price depending on experience.
Stack: Python, FastAPI, Supabase/PostgreSQL, Claude API.

If anyone on the team does consulting, or can point me to someone who's implemented pyBKT in production, I'd appreciate it.

Erik Raschke
erik@shryn.ai

[energypantry]

Hi Erik — I can help with this.

I can integrate pyBKT into your Python/FastAPI + Postgres stack (Supabase), add tests, and expose clean per-skill mastery + overall pass-probability outputs.

For the non-binary 0–100 Claude scores, two practical approaches that still fit “BKT-style” production needs:

1. Soft-evidence update: treat the score as P(correct) (fractional credit) via a calibrated mapping, then update the posterior accordingly.
2. Binning/threshold + per-skill calibration: discretize to a small number of bins (or binary) and fit slip/guess via EM on your historical data.

5-day sprint proposal:

• Day 1: review spec/code + confirm skill tagging + define evaluation metric + sample data pull
• Day 2–3: baseline BKT integration + endpoints + tests
• Day 4: continuous-score extension + calibration/backtest
• Day 5: cleanup, docs, deployment checklist

If you’re still hiring, can you share:

• how skills are defined/tagged
• a small anonymized sample (student_id, skill_id, score, timestamp)
• how you want “pass probability” computed/validated

Budget: milestone-based fixed price within your $500–$1,500 range. Payment-wise PayPal is fine; USDC/USDT also works.

Reply here if you want to proceed and I’ll send contact details.

[energypantry]

Hi Erik — quick follow-up to see if you're still looking for help on this BKT production integration.

If yes, I can start this week. To scope it cleanly, can you share:

• how skills are defined/tagged today
• • a small anonymized sample (student_id, skill_id, score, timestamp)
• • • what “pass probability” should mean in your product (definition + how you’ll validate it)
      Once I see those, I’ll propose a tight milestone plan + fixed-price quote within your $500–$1,500 range.

Thanks.

[energypantry]

Quick follow-up: I can start on this immediately.

If you share (1) a small sample of the raw data/events you want the BKT to run on, and (2) the expected output format (and any constraints around latency/throughput), I can send back a concrete 5-day plan + milestones.

Happy to work async here or jump on a short call if that is easier.

[energypantry]

Following up in case this consulting sprint is still open.

I can start with a focused 5-day delivery plan (integration + validation + handoff notes), and adapt scope to your preferred budget band.

If it is still active, I can begin immediately after you share the current dataset/interface constraints.

[nikhilbalwani]

@energypantry Been following this conversation. If they are interested, they will reach out. If they are ignoring you, they are doing it on purpose. Don't force it :) Please don't feel compelled to follow up multiple times, you are meant for better. Besides, it is not even a lot of money for the amount of work expected.